Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Found mnist Data
Found celeba Data

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x7f5e14c60c18>

# import matplotlib
# matplotlib.__version__
# !pip install -U matplotlib==2.0.2

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7f5e14b58208>

\\\\\\\\\\\\\\\\\\\\\\## Preprocess the Data Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf
import tensorflow.contrib.slim as slim

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

/home/jsingh/anaconda3/envs/cv3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters

WARNING:tensorflow:From /home/jsingh/anaconda3/envs/cv3/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/datasets/base.py:198: retry (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.
Instructions for updating:
Use the retry module or similar alternatives.
TensorFlow Version: 1.7.0
Default GPU Device: /device:GPU:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    # TODO: Implement Function
    input_real = tf.placeholder(tf.float32,[None,image_width,image_height,image_channels])
    input_z = tf.placeholder(tf.float32,[None,z_dim])
    learning_rate = tf.placeholder(tf.float32)
    return input_real,input_z,learning_rate


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

def discriminator(images, reuse=False,alpha=0.2):
    """
    Create the discriminator network
    :param images: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    # TODO: Implement Function
    with tf.variable_scope("discriminator",reuse=reuse):
        # input_shape : (32,32,3)
        net = slim.conv2d(images,64,5,stride=2,activation_fn=None) # (32,32,16)
        net = tf.maximum(alpha*net,net)
        net = slim.conv2d(net,128,5,stride=2,activation_fn=None)
        net = slim.batch_norm(net)
        net = tf.maximum(alpha*net,net)
        net = slim.conv2d(net,256,5,stride=2,activation_fn=None)
        net = slim.batch_norm(net)
        net = tf.maximum(alpha*net,net)
        net = slim.flatten(net)
        logits = slim.fully_connected(net,1,activation_fn=None)
        out = tf.sigmoid(logits)
    return out,logits


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z,out_channel_dim,is_train=True,alpha=.2):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
    
    # TODO: Implement Function
    with tf.variable_scope("generator",reuse=not is_train):
        net = slim.fully_connected(z,4*4*512,activation_fn=None)
        net = tf.reshape(net,[-1,4,4,512])
        net = slim.batch_norm(net,is_training=is_train)
        net = tf.maximum(alpha*net,net)    # (4,4,)
        net = tf.layers.conv2d_transpose(net,256,5,strides=2)
        net = slim.batch_norm(net,is_training=is_train)
        net = tf.maximum(alpha*net,net)    # (8,8,)
        net = tf.layers.conv2d_transpose(net,128,5,strides=2)
        net = slim.batch_norm(net,is_training=is_train)         
        net = tf.maximum(alpha*net,net)    # (16,16,)

        net = tf.layers.conv2d_transpose(net,out_channel_dim,5,strides=2) #(32,32,output_channel_dim)
        net = tf.image.resize_images(net,[28,28]) # (28,28,C)

        out = .5*tf.tanh(net)
    return out


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""

tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    # TODO: Implement Function
    gen_out = generator(input_z,out_channel_dim,is_train=True)
    d_out_real,d_logits_real = discriminator(input_real,reuse=False)
    d_out_fake,d_logits_fake = discriminator(gen_out,reuse=True)
    d_loss_real = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.ones_like(d_logits_real)
                                                          ,logits=d_logits_real))
    d_loss_fake = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.zeros_like(d_logits_fake)
                                                          ,logits=d_logits_fake))
    d_loss = d_loss_real+d_loss_fake
    g_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.ones_like(d_logits_fake)
                                                     ,logits=d_logits_fake))
    
    return d_loss, g_loss


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    # TODO: Implement Function
    # Get weights and bias to update
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]
    
   # Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)

    return d_train_opt, g_train_opt


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    
    # print parameters
    DISPLAY_STEP = 100
    GEN_DISPLAY_STEP = 10*DISPLAY_STEP
    
    # TODO: Build Model
    step = 0
    im_width , im_height , im_channels = data_shape[1:]
    
    input_real,input_z,lr = model_inputs(im_width,im_height,im_channels,z_dim)
    d_loss,g_loss = model_loss(input_real,input_z,im_channels)
    d_opt,g_opt = model_opt(d_loss,g_loss,lr,beta1)
     
    with tf.Session() as sess:    
        sess.run([tf.global_variables_initializer(),tf.local_variables_initializer()])
        for epoch_i in range(epoch_count):
            for batch_images in get_batches(batch_size):
                step += 1 
                # TODO: Train Model
                
                # getting batch data
                batch_real = batch_images
                
                # Sample random noise for G
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))
                
                # training the model
                _,d_l = sess.run([d_opt,d_loss],feed_dict={input_real:batch_real,input_z:batch_z,
                                                           lr:learning_rate})
                _,g_l = sess.run([g_opt,g_loss],feed_dict={input_real:batch_real,input_z:batch_z
                                                           ,lr:learning_rate})
                
                if step % DISPLAY_STEP == 0 :
                    print ("Epoch : {}/{} ... Step : {} ... d_loss : {} ... g_loss : {}"
                           .format(epoch_i+1,epoch_count,step,d_l,g_l))
                
                if step % GEN_DISPLAY_STEP == 0 :
                    n_images = 25
                    show_generator_output(sess,n_images,input_z,im_channels,data_image_mode)

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 64
z_dim = 100
learning_rate = 3e-3
beta1 = .5

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 10

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch : 1/10 ... Step : 10 ... d_loss : 0.9336976408958435 ... g_loss : 7.311212539672852
Epoch : 1/10 ... Step : 20 ... d_loss : 0.3639373183250427 ... g_loss : 3.2976484298706055
Epoch : 1/10 ... Step : 30 ... d_loss : 0.17301833629608154 ... g_loss : 4.143669128417969
Epoch : 1/10 ... Step : 40 ... d_loss : 0.013400103896856308 ... g_loss : 6.166452407836914
Epoch : 1/10 ... Step : 50 ... d_loss : 0.029078524559736252 ... g_loss : 7.749451637268066
Epoch : 1/10 ... Step : 60 ... d_loss : 1.6513876914978027 ... g_loss : 9.911873817443848
Epoch : 1/10 ... Step : 70 ... d_loss : 0.1284271776676178 ... g_loss : 12.684028625488281
Epoch : 1/10 ... Step : 80 ... d_loss : 7.0270915031433105 ... g_loss : 4.399494647979736
Epoch : 1/10 ... Step : 90 ... d_loss : 0.1814710795879364 ... g_loss : 6.094439506530762
Epoch : 1/10 ... Step : 100 ... d_loss : 1.3318357467651367 ... g_loss : 0.6794193983078003

Epoch : 1/10 ... Step : 110 ... d_loss : 1.397844910621643 ... g_loss : 3.883117198944092
Epoch : 1/10 ... Step : 120 ... d_loss : 1.729637861251831 ... g_loss : 0.9159450531005859
Epoch : 1/10 ... Step : 130 ... d_loss : 0.9159771203994751 ... g_loss : 2.8071227073669434
Epoch : 1/10 ... Step : 140 ... d_loss : 0.7850979566574097 ... g_loss : 2.634237289428711
Epoch : 1/10 ... Step : 150 ... d_loss : 0.7400979995727539 ... g_loss : 3.4742345809936523
Epoch : 1/10 ... Step : 160 ... d_loss : 1.6995909214019775 ... g_loss : 0.6127923727035522
Epoch : 1/10 ... Step : 170 ... d_loss : 1.0672526359558105 ... g_loss : 2.419649600982666
Epoch : 1/10 ... Step : 180 ... d_loss : 2.686013698577881 ... g_loss : 0.673509955406189
Epoch : 1/10 ... Step : 190 ... d_loss : 1.4837532043457031 ... g_loss : 4.456912040710449
Epoch : 1/10 ... Step : 200 ... d_loss : 0.5718165636062622 ... g_loss : 3.2093491554260254

Epoch : 1/10 ... Step : 210 ... d_loss : 1.0083123445510864 ... g_loss : 2.3459014892578125
Epoch : 1/10 ... Step : 220 ... d_loss : 2.623124361038208 ... g_loss : 1.6884279251098633
Epoch : 1/10 ... Step : 230 ... d_loss : 1.198453426361084 ... g_loss : 2.7551279067993164
Epoch : 1/10 ... Step : 240 ... d_loss : 0.45371878147125244 ... g_loss : 3.0702357292175293
Epoch : 1/10 ... Step : 250 ... d_loss : 1.5990288257598877 ... g_loss : 1.3336373567581177
Epoch : 1/10 ... Step : 260 ... d_loss : 0.5275869369506836 ... g_loss : 2.115786552429199
Epoch : 1/10 ... Step : 270 ... d_loss : 1.4636509418487549 ... g_loss : 3.59511661529541
Epoch : 1/10 ... Step : 280 ... d_loss : 0.9516904354095459 ... g_loss : 3.3252086639404297
Epoch : 1/10 ... Step : 290 ... d_loss : 0.7114964723587036 ... g_loss : 3.264331817626953
Epoch : 1/10 ... Step : 300 ... d_loss : 0.8348485231399536 ... g_loss : 2.270519256591797

Epoch : 1/10 ... Step : 310 ... d_loss : 1.294409155845642 ... g_loss : 1.336060881614685
Epoch : 1/10 ... Step : 320 ... d_loss : 0.921809196472168 ... g_loss : 5.333911895751953
Epoch : 1/10 ... Step : 330 ... d_loss : 0.895397424697876 ... g_loss : 2.6435799598693848
Epoch : 1/10 ... Step : 340 ... d_loss : 0.8738511204719543 ... g_loss : 3.386038064956665
Epoch : 1/10 ... Step : 350 ... d_loss : 0.620734691619873 ... g_loss : 3.7486414909362793
Epoch : 1/10 ... Step : 360 ... d_loss : 1.0144203901290894 ... g_loss : 1.469957947731018
Epoch : 1/10 ... Step : 370 ... d_loss : 0.39007750153541565 ... g_loss : 3.1703720092773438
Epoch : 1/10 ... Step : 380 ... d_loss : 1.1083438396453857 ... g_loss : 1.4196758270263672
Epoch : 1/10 ... Step : 390 ... d_loss : 0.8865296840667725 ... g_loss : 1.097640872001648
Epoch : 1/10 ... Step : 400 ... d_loss : 1.1189064979553223 ... g_loss : 1.8318194150924683

Epoch : 1/10 ... Step : 410 ... d_loss : 1.3792486190795898 ... g_loss : 3.445786952972412
Epoch : 1/10 ... Step : 420 ... d_loss : 1.1093313694000244 ... g_loss : 2.5241875648498535
Epoch : 1/10 ... Step : 430 ... d_loss : 0.9715054035186768 ... g_loss : 3.209750175476074
Epoch : 1/10 ... Step : 440 ... d_loss : 1.544184684753418 ... g_loss : 5.252941608428955
Epoch : 1/10 ... Step : 450 ... d_loss : 0.4542277455329895 ... g_loss : 2.635376453399658
Epoch : 1/10 ... Step : 460 ... d_loss : 1.1781299114227295 ... g_loss : 3.10211443901062
Epoch : 1/10 ... Step : 470 ... d_loss : 1.722873330116272 ... g_loss : 0.3512219488620758
Epoch : 1/10 ... Step : 480 ... d_loss : 1.0127923488616943 ... g_loss : 2.341762065887451
Epoch : 1/10 ... Step : 490 ... d_loss : 1.004692554473877 ... g_loss : 2.802274227142334
Epoch : 1/10 ... Step : 500 ... d_loss : 1.08346426486969 ... g_loss : 1.052732229232788

Epoch : 1/10 ... Step : 510 ... d_loss : 0.8115843534469604 ... g_loss : 2.279780864715576
Epoch : 1/10 ... Step : 520 ... d_loss : 1.4663351774215698 ... g_loss : 0.38215139508247375
Epoch : 1/10 ... Step : 530 ... d_loss : 0.8898907899856567 ... g_loss : 2.384796380996704
Epoch : 1/10 ... Step : 540 ... d_loss : 1.2728767395019531 ... g_loss : 1.1691789627075195
Epoch : 1/10 ... Step : 550 ... d_loss : 0.8470571041107178 ... g_loss : 2.0264289379119873
Epoch : 1/10 ... Step : 560 ... d_loss : 1.5382022857666016 ... g_loss : 1.50447678565979
Epoch : 1/10 ... Step : 570 ... d_loss : 1.6164087057113647 ... g_loss : 0.7648814916610718
Epoch : 1/10 ... Step : 580 ... d_loss : 1.1318968534469604 ... g_loss : 1.7330741882324219
Epoch : 1/10 ... Step : 590 ... d_loss : 1.6620967388153076 ... g_loss : 0.4128282070159912
Epoch : 1/10 ... Step : 600 ... d_loss : 1.0595319271087646 ... g_loss : 1.251063346862793

Epoch : 1/10 ... Step : 610 ... d_loss : 1.7486302852630615 ... g_loss : 0.644080400466919
Epoch : 1/10 ... Step : 620 ... d_loss : 1.0599135160446167 ... g_loss : 1.5116643905639648
Epoch : 1/10 ... Step : 630 ... d_loss : 1.2168996334075928 ... g_loss : 0.9227011203765869
Epoch : 1/10 ... Step : 640 ... d_loss : 0.8925243020057678 ... g_loss : 1.7199068069458008
Epoch : 1/10 ... Step : 650 ... d_loss : 1.147553563117981 ... g_loss : 1.5625324249267578
Epoch : 1/10 ... Step : 660 ... d_loss : 0.9455957412719727 ... g_loss : 1.318260669708252
Epoch : 1/10 ... Step : 670 ... d_loss : 1.1988613605499268 ... g_loss : 2.234072685241699
Epoch : 1/10 ... Step : 680 ... d_loss : 0.8029910326004028 ... g_loss : 2.58685302734375
Epoch : 1/10 ... Step : 690 ... d_loss : 0.9784906506538391 ... g_loss : 2.1536436080932617
Epoch : 1/10 ... Step : 700 ... d_loss : 0.847655177116394 ... g_loss : 2.493992328643799

Epoch : 1/10 ... Step : 710 ... d_loss : 1.828971266746521 ... g_loss : 0.4550655782222748
Epoch : 1/10 ... Step : 720 ... d_loss : 1.6097397804260254 ... g_loss : 0.8204618692398071
Epoch : 1/10 ... Step : 730 ... d_loss : 1.0887187719345093 ... g_loss : 2.7606587409973145
Epoch : 1/10 ... Step : 740 ... d_loss : 0.963829517364502 ... g_loss : 1.1489312648773193
Epoch : 1/10 ... Step : 750 ... d_loss : 0.8147303462028503 ... g_loss : 1.6449116468429565
Epoch : 1/10 ... Step : 760 ... d_loss : 1.2644479274749756 ... g_loss : 1.245898962020874
Epoch : 1/10 ... Step : 770 ... d_loss : 2.3815462589263916 ... g_loss : 0.7054839730262756
Epoch : 1/10 ... Step : 780 ... d_loss : 1.0985352993011475 ... g_loss : 1.199989914894104
Epoch : 1/10 ... Step : 790 ... d_loss : 0.8132576942443848 ... g_loss : 2.0180912017822266
Epoch : 1/10 ... Step : 800 ... d_loss : 1.3107726573944092 ... g_loss : 2.2561357021331787

Epoch : 1/10 ... Step : 810 ... d_loss : 1.0638477802276611 ... g_loss : 1.8573417663574219
Epoch : 1/10 ... Step : 820 ... d_loss : 1.1857035160064697 ... g_loss : 0.7945778369903564
Epoch : 1/10 ... Step : 830 ... d_loss : 1.0588810443878174 ... g_loss : 1.7995920181274414
Epoch : 1/10 ... Step : 840 ... d_loss : 0.7991496920585632 ... g_loss : 2.8622632026672363
Epoch : 1/10 ... Step : 850 ... d_loss : 1.6774568557739258 ... g_loss : 0.7666798233985901
Epoch : 1/10 ... Step : 860 ... d_loss : 1.1381988525390625 ... g_loss : 1.804802656173706
Epoch : 1/10 ... Step : 870 ... d_loss : 1.1858789920806885 ... g_loss : 1.1012569665908813
Epoch : 1/10 ... Step : 880 ... d_loss : 1.517936110496521 ... g_loss : 1.2802445888519287
Epoch : 1/10 ... Step : 890 ... d_loss : 0.6709963083267212 ... g_loss : 2.9474101066589355
Epoch : 1/10 ... Step : 900 ... d_loss : 1.611269474029541 ... g_loss : 0.9779760837554932

Epoch : 1/10 ... Step : 910 ... d_loss : 0.7024667263031006 ... g_loss : 2.1487045288085938
Epoch : 1/10 ... Step : 920 ... d_loss : 2.2447454929351807 ... g_loss : 1.576002836227417
Epoch : 1/10 ... Step : 930 ... d_loss : 0.914207398891449 ... g_loss : 4.081455230712891
Epoch : 2/10 ... Step : 940 ... d_loss : 1.1459710597991943 ... g_loss : 1.118256688117981
Epoch : 2/10 ... Step : 950 ... d_loss : 0.8255268335342407 ... g_loss : 2.19569730758667
Epoch : 2/10 ... Step : 960 ... d_loss : 0.8839629292488098 ... g_loss : 2.1204190254211426
Epoch : 2/10 ... Step : 970 ... d_loss : 1.1559247970581055 ... g_loss : 3.868286609649658
Epoch : 2/10 ... Step : 980 ... d_loss : 0.7506850957870483 ... g_loss : 1.4702450037002563
Epoch : 2/10 ... Step : 990 ... d_loss : 1.1529264450073242 ... g_loss : 3.8402624130249023
Epoch : 2/10 ... Step : 1000 ... d_loss : 1.6934853792190552 ... g_loss : 0.965009331703186

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 64
z_dim = 200
learning_rate = 1e-3
beta1 = .5


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 10

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch : 1/10 ... Step : 100 ... d_loss : 0.276090532541275 ... g_loss : 7.946306228637695
Epoch : 1/10 ... Step : 200 ... d_loss : 0.4930230379104614 ... g_loss : 7.372992515563965
Epoch : 1/10 ... Step : 300 ... d_loss : 0.5517822504043579 ... g_loss : 4.447853088378906
Epoch : 1/10 ... Step : 400 ... d_loss : 0.7986226081848145 ... g_loss : 2.0185770988464355
Epoch : 1/10 ... Step : 500 ... d_loss : 1.2721617221832275 ... g_loss : 2.4557392597198486
Epoch : 1/10 ... Step : 600 ... d_loss : 0.9766598343849182 ... g_loss : 2.350954055786133
Epoch : 1/10 ... Step : 700 ... d_loss : 1.061157464981079 ... g_loss : 2.4358301162719727
Epoch : 1/10 ... Step : 800 ... d_loss : 1.1672296524047852 ... g_loss : 1.8027163743972778
Epoch : 1/10 ... Step : 900 ... d_loss : 1.6221281290054321 ... g_loss : 3.5671944618225098
Epoch : 1/10 ... Step : 1000 ... d_loss : 1.0373295545578003 ... g_loss : 2.998807907104492

Epoch : 1/10 ... Step : 1100 ... d_loss : 1.1935968399047852 ... g_loss : 1.9688359498977661
Epoch : 1/10 ... Step : 1200 ... d_loss : 1.2537930011749268 ... g_loss : 1.2126582860946655
Epoch : 1/10 ... Step : 1300 ... d_loss : 0.9132339954376221 ... g_loss : 2.275480270385742
Epoch : 1/10 ... Step : 1400 ... d_loss : 1.2980901002883911 ... g_loss : 3.1403751373291016
Epoch : 1/10 ... Step : 1500 ... d_loss : 0.9228848218917847 ... g_loss : 1.241837978363037
Epoch : 1/10 ... Step : 1600 ... d_loss : 1.315888524055481 ... g_loss : 2.451977014541626
Epoch : 1/10 ... Step : 1700 ... d_loss : 0.9177507758140564 ... g_loss : 3.095423698425293
Epoch : 1/10 ... Step : 1800 ... d_loss : 1.277606725692749 ... g_loss : 1.3688812255859375
Epoch : 1/10 ... Step : 1900 ... d_loss : 0.986170768737793 ... g_loss : 1.9964361190795898
Epoch : 1/10 ... Step : 2000 ... d_loss : 1.0518923997879028 ... g_loss : 2.5830845832824707

Epoch : 1/10 ... Step : 2100 ... d_loss : 0.980575442314148 ... g_loss : 1.716047763824463
Epoch : 1/10 ... Step : 2200 ... d_loss : 0.7019271850585938 ... g_loss : 1.8456089496612549
Epoch : 1/10 ... Step : 2300 ... d_loss : 0.7612940073013306 ... g_loss : 2.375460386276245
Epoch : 1/10 ... Step : 2400 ... d_loss : 1.142967700958252 ... g_loss : 1.5365822315216064
Epoch : 1/10 ... Step : 2500 ... d_loss : 0.9833749532699585 ... g_loss : 1.5736494064331055
Epoch : 1/10 ... Step : 2600 ... d_loss : 1.3212558031082153 ... g_loss : 1.7395437955856323
Epoch : 1/10 ... Step : 2700 ... d_loss : 1.0128319263458252 ... g_loss : 1.5958023071289062
Epoch : 1/10 ... Step : 2800 ... d_loss : 1.1938549280166626 ... g_loss : 1.3919132947921753
Epoch : 1/10 ... Step : 2900 ... d_loss : 1.1105329990386963 ... g_loss : 1.8811490535736084
Epoch : 1/10 ... Step : 3000 ... d_loss : 1.0103437900543213 ... g_loss : 1.3165602684020996

Epoch : 1/10 ... Step : 3100 ... d_loss : 0.8804258108139038 ... g_loss : 1.5781946182250977
Epoch : 2/10 ... Step : 3200 ... d_loss : 1.0533196926116943 ... g_loss : 2.0449299812316895
Epoch : 2/10 ... Step : 3300 ... d_loss : 1.0396111011505127 ... g_loss : 1.7734143733978271
Epoch : 2/10 ... Step : 3400 ... d_loss : 1.216396450996399 ... g_loss : 1.3083497285842896
Epoch : 2/10 ... Step : 3500 ... d_loss : 0.8309692144393921 ... g_loss : 2.2075929641723633
Epoch : 2/10 ... Step : 3600 ... d_loss : 0.9828373193740845 ... g_loss : 1.773605465888977
Epoch : 2/10 ... Step : 3700 ... d_loss : 0.9848295450210571 ... g_loss : 3.1024365425109863
Epoch : 2/10 ... Step : 3800 ... d_loss : 0.9401560425758362 ... g_loss : 2.2019243240356445
Epoch : 2/10 ... Step : 3900 ... d_loss : 1.4543629884719849 ... g_loss : 1.047347068786621
Epoch : 2/10 ... Step : 4000 ... d_loss : 0.7480818629264832 ... g_loss : 2.766585350036621

Epoch : 2/10 ... Step : 4100 ... d_loss : 1.1502654552459717 ... g_loss : 1.4453022480010986
Epoch : 2/10 ... Step : 4200 ... d_loss : 1.403666377067566 ... g_loss : 3.304753541946411
Epoch : 2/10 ... Step : 4300 ... d_loss : 1.026075839996338 ... g_loss : 1.3864378929138184
Epoch : 2/10 ... Step : 4400 ... d_loss : 1.0760143995285034 ... g_loss : 2.1319215297698975
Epoch : 2/10 ... Step : 4500 ... d_loss : 1.3108898401260376 ... g_loss : 2.742766857147217
Epoch : 2/10 ... Step : 4600 ... d_loss : 1.506540060043335 ... g_loss : 1.1960229873657227
Epoch : 2/10 ... Step : 4700 ... d_loss : 1.1399314403533936 ... g_loss : 4.008848667144775
Epoch : 2/10 ... Step : 4800 ... d_loss : 0.7025111317634583 ... g_loss : 1.8710358142852783
Epoch : 2/10 ... Step : 4900 ... d_loss : 1.1308298110961914 ... g_loss : 2.3569982051849365
Epoch : 2/10 ... Step : 5000 ... d_loss : 0.6579766273498535 ... g_loss : 2.144453525543213

Epoch : 2/10 ... Step : 5100 ... d_loss : 0.794979989528656 ... g_loss : 2.1908798217773438
Epoch : 2/10 ... Step : 5200 ... d_loss : 0.7883226871490479 ... g_loss : 2.3531274795532227
Epoch : 2/10 ... Step : 5300 ... d_loss : 1.1059893369674683 ... g_loss : 2.5480332374572754
Epoch : 2/10 ... Step : 5400 ... d_loss : 0.8394128680229187 ... g_loss : 2.898387908935547
Epoch : 2/10 ... Step : 5500 ... d_loss : 0.6525368690490723 ... g_loss : 1.7932838201522827
Epoch : 2/10 ... Step : 5600 ... d_loss : 0.9933221340179443 ... g_loss : 2.732285976409912
Epoch : 2/10 ... Step : 5700 ... d_loss : 1.815030574798584 ... g_loss : 0.6647165417671204
Epoch : 2/10 ... Step : 5800 ... d_loss : 0.3374061584472656 ... g_loss : 3.555860996246338
Epoch : 2/10 ... Step : 5900 ... d_loss : 1.0418013334274292 ... g_loss : 0.9163437485694885
Epoch : 2/10 ... Step : 6000 ... d_loss : 0.9274795055389404 ... g_loss : 1.5534119606018066

Epoch : 2/10 ... Step : 6100 ... d_loss : 1.0513306856155396 ... g_loss : 2.5936737060546875
Epoch : 2/10 ... Step : 6200 ... d_loss : 0.7317748069763184 ... g_loss : 3.2648637294769287
Epoch : 2/10 ... Step : 6300 ... d_loss : 0.9945372343063354 ... g_loss : 1.7116374969482422
Epoch : 3/10 ... Step : 6400 ... d_loss : 0.6944642066955566 ... g_loss : 2.5919370651245117
Epoch : 3/10 ... Step : 6500 ... d_loss : 1.3452577590942383 ... g_loss : 4.543449401855469
Epoch : 3/10 ... Step : 6600 ... d_loss : 0.6878675222396851 ... g_loss : 2.296178102493286
Epoch : 3/10 ... Step : 6700 ... d_loss : 0.813028872013092 ... g_loss : 2.1186952590942383
Epoch : 3/10 ... Step : 6800 ... d_loss : 0.7233805656433105 ... g_loss : 2.300830364227295
Epoch : 3/10 ... Step : 6900 ... d_loss : 0.7770169377326965 ... g_loss : 2.75323486328125
Epoch : 3/10 ... Step : 7000 ... d_loss : 1.0308586359024048 ... g_loss : 1.4336400032043457

Epoch : 3/10 ... Step : 7100 ... d_loss : 1.5989021062850952 ... g_loss : 0.7926545143127441
Epoch : 3/10 ... Step : 7200 ... d_loss : 0.8878282904624939 ... g_loss : 2.640322685241699
Epoch : 3/10 ... Step : 7300 ... d_loss : 0.841025173664093 ... g_loss : 1.9909579753875732
Epoch : 3/10 ... Step : 7400 ... d_loss : 0.7161791324615479 ... g_loss : 3.7084102630615234
Epoch : 3/10 ... Step : 7500 ... d_loss : 0.9519063234329224 ... g_loss : 1.9285218715667725
Epoch : 3/10 ... Step : 7600 ... d_loss : 0.6772498488426208 ... g_loss : 3.6724724769592285
Epoch : 3/10 ... Step : 7700 ... d_loss : 1.0046195983886719 ... g_loss : 1.606713056564331
Epoch : 3/10 ... Step : 7800 ... d_loss : 0.7340472936630249 ... g_loss : 2.3877339363098145
Epoch : 3/10 ... Step : 7900 ... d_loss : 0.6103564500808716 ... g_loss : 2.388050079345703
Epoch : 3/10 ... Step : 8000 ... d_loss : 1.90255606174469 ... g_loss : 4.127630710601807

Epoch : 3/10 ... Step : 8100 ... d_loss : 0.8253073692321777 ... g_loss : 1.6821367740631104
Epoch : 3/10 ... Step : 8200 ... d_loss : 1.077162504196167 ... g_loss : 2.130045175552368
Epoch : 3/10 ... Step : 8300 ... d_loss : 0.8370654582977295 ... g_loss : 1.717990517616272
Epoch : 3/10 ... Step : 8400 ... d_loss : 0.9179205894470215 ... g_loss : 3.027897357940674
Epoch : 3/10 ... Step : 8500 ... d_loss : 0.7762618660926819 ... g_loss : 3.4756534099578857
Epoch : 3/10 ... Step : 8600 ... d_loss : 0.8043072819709778 ... g_loss : 1.7576829195022583
Epoch : 3/10 ... Step : 8700 ... d_loss : 0.8062787055969238 ... g_loss : 1.5709490776062012
Epoch : 3/10 ... Step : 8800 ... d_loss : 0.7743768095970154 ... g_loss : 2.53987979888916
Epoch : 3/10 ... Step : 8900 ... d_loss : 0.5028189420700073 ... g_loss : 2.766352653503418
Epoch : 3/10 ... Step : 9000 ... d_loss : 0.9073359966278076 ... g_loss : 3.59749698638916

Epoch : 3/10 ... Step : 9100 ... d_loss : 0.7329654097557068 ... g_loss : 2.858391523361206
Epoch : 3/10 ... Step : 9200 ... d_loss : 0.6823814511299133 ... g_loss : 3.1438064575195312
Epoch : 3/10 ... Step : 9300 ... d_loss : 0.882191002368927 ... g_loss : 3.844712495803833
Epoch : 3/10 ... Step : 9400 ... d_loss : 0.4378136694431305 ... g_loss : 2.3995633125305176
Epoch : 4/10 ... Step : 9500 ... d_loss : 0.8612440228462219 ... g_loss : 2.706346273422241
Epoch : 4/10 ... Step : 9600 ... d_loss : 1.3807734251022339 ... g_loss : 1.6906564235687256
Epoch : 4/10 ... Step : 9700 ... d_loss : 1.057225227355957 ... g_loss : 1.6074702739715576
Epoch : 4/10 ... Step : 9800 ... d_loss : 0.47769269347190857 ... g_loss : 3.713808536529541
Epoch : 4/10 ... Step : 9900 ... d_loss : 1.0058733224868774 ... g_loss : 1.3367810249328613
Epoch : 4/10 ... Step : 10000 ... d_loss : 1.397682547569275 ... g_loss : 1.5888102054595947

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.